Powerful tools, these require some adjustments. Their goal, deepen research and respond more relevantly.
The AI extends its powers with the connectors. At the start of the year, it was Openai’s turn to present it for his Deep Research. As their name suggests, they allow you to connect to various services. The goal is to allow Chatgpt to have directly access to data. The conversational agent can rely on data to respond more effectively to prompt. “Use connectors in Deep Research to generate long and sourced answers integrating the internal tools of your business,” writes the San Francisco company.
The functioning of connectors is in theory quite simple. Go to the Chatgpt interface. Click Tools, and “In -depth research”. Indeed, to use the connectors, you must go through Deep Research. This agent, who uses the O3 model, is able to conduct complex research independently. Traveling in navigation and reasoning tasks, he browsing websites and interpreting the content, PDF and images tables. It can treat hundreds of different sources. Since February 25, 2025, it has been available for Chatgpt Plus subscribers, Team, Edu and Enterprise. Then go to “Sources” and to the connector concerned. You can then ask your question in the tool.
Github to help developers and neophytes
Chatgpt supports advanced search in Gmail, Calendar, Drive and Github. We start by testing Github with Chatgpt Plus. Launched on May 8, 2025, the GitHub connector offers direct access to the deposits present on the accommodation and software development management service. It allows you to explore and question code content on Chatgpt directly. You can ask him what is in a deposit, or even create new elements. The tool can notably offer steps step by step to move forward in your project.
To use it, we will connect our Github account, where at least one deposit must appear. The connection to it can be delicate. Multifactor authentication (MFA) must provide at least two verification factors. The deposit can put ten minutes before being indexed by Chatgpt. A number of problems can intervene. If your deposit is not indexed by GitHub, you can manually trigger indexing on GitHub thanks to a search for the deposit in the following format: repo: {username/repo_name} import (for example, repo: OPENAI/Import codex). If your repository is private or created after a first connection to GitHub, you can go through the “parameters” of Chatgpt, at the bottom right of the interface, “connectors”, “github” and “manage” to solve the problem. We advise you, if possible, to pass the deposit in public mode.
Once the deposit has been connected, we launch a relatively simple prompt:
“Within nomicatorgithub/directory tell me what is in the” PKG “file. Give me the tree structure with the different files present. Make a brief description of each file.” After a few minutes, Deep Research gives his answer.
We see in the extract a detailed, structured and precise response. Be careful to check the answers given. Links allow you to go directly to GitHub on the requested page. It is also easy to ask Chatgpt to correct, or to add an element. Non -developers can, for example, request help to understand these documents.
On the confidentiality side, note that only shared data are accessible. This is an advantage for sensitive files. If you use the version Plus or Pro, you can deactivate the “Improve the model for all” option to prevent the American company from using your content to train its models.
Google Agenda to plan or receive briefs
For its part, the Google Calendar connector allows you to add events, plan, or even integrate events into conversations with chatgpt, in the form of a recall for example. You can also get suggestions for meeting slots, or even receive briefs of the day every morning on the next meetings, with notes for each of them.
The installation of this connector is easy. Just go to the chatgpt interface, go to in -depth search and activate the Google Agenda connector. Then confirm the authorizations on the Google page.
As part of one of our tests, we submit this prompt to him: “Add me an appointment on my agenda on Tuesday July 29 at 3 p.m. French time. This is a 1/2 hour video with a company with a Google Meet link. Adds a videoconferencing link. The other participant in this meeting is at contact@entreprise.fr. The title of the meeting is” Rendezen Visio Entreprise “.
Chatgpt requests the possibility of adding an automatic reminder.
After our answer, we see on this video extract the chatgpt paths. This one seeks for example to create an event on Google Calendar “using Python and an API, while respecting the date and time formats and the constraints of the system.” It structures the event. The answer takes approximately 5 minutes to come. The result is interesting, with a sheet, the date, the time, the duration, the title of the event, the link, the description and the programmed recall. Unfortunately, according to our different tests, the link is not displayed in our Google agenda. This problem should be corrected quickly.
On the security side, let us indicate that less than submit your data to OpenAi as part of a feedback, the American company declares that it does not cause its generalized models from this data or information derived from this data.
Gmail to analyze relationships
Another connector offered by Deep Research of Chatgpt: Gmail. The connection is significantly the same as that of Google Agenda. The connector is mainly used for in -depth tasks. It is particularly useful for detailed reports and contact analyzes. Example of possibility for certain companies, the possibility of analyzing the evolution of the relationship with a customer. According to our tests, Chatgpt can produce a clear calendar of the relationship and suggest the next strategic steps.
For example, we asked him to analyze the evolution of the relationship with a company, focusing on changes in feelings. We do not broadcast the answer, as a confidentiality.
The result is quite stunning. Chatgpt provides “global” synthesis. By relying on written language, through tone, expressiveness, type of communication, but also the frequency of shipments, speed of response to messages, or the different events that have impacted the relationship, the tool offers a relatively fine and nuanced reading. His answer appears in the form of a chronological analysis. However, be careful to check the accuracy of the elements provided.
